The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs and executes shell commands for various platform-specific CLI tools (GitHub's gh, GitLab's glab, Azure's az, Gitea's tea) and uses curl for Bitbucket APIs. It populates command arguments such as --title and --body (or --description) using data extracted directly from the git repository.
Evidence: In "Paso 7 — Crear PR/MR", the skill defines commands like gh pr create --title "<título>" --body "<desc>". These variables are derived from git log and branch names in "Paso 6".
Risk: If a malicious contributor provides a commit message or branch name containing shell metacharacters (e.g., backticks, command substitution strings like $(...), or pipes), it could lead to arbitrary command execution on the user's machine when the CLI tool is invoked via the shell.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the git environment which could be used to host malicious instructions.
Ingestion points: The skill reads data from git log, git diff, and the contents of the docs/policies/definition-of-done.md file (Step 4.4).
Boundary markers: No specific boundary markers or instructions to "ignore embedded commands" are used when the agent processes the contents of the git metadata or the policy files.
Capability inventory: The skill can execute various shell commands, perform network operations via curl, and modify files in the docs/audits/ directory.
Sanitization: There is no evidence of sanitization or escaping of the ingested git data or file contents before they are interpolated into the PR metadata or used for verification logic.
[DYNAMIC_EXECUTION]: The skill dynamically selects and assembles command-line strings at runtime based on the detected git remote platform.
Evidence: The "Detección de plataforma" table and Step 7 show the skill switching its execution logic and command construction based on the git remote get-url origin output.

pr-create